I just ripped a stack of about 40 CDs accumulated during trips to Goodwill. Almost half failed Musicbrainz lookup (mostly expected because they're obscure, but some surprising). So I spent a fair amount of time keying in info about them. Because I didn't RTFM thoroughly enough, and had set the configuration to use Musicbrainz, I ASSumed that my entries would be automatically uploaded. Oops. I can kinda-sorta recover from my mistake by writing a script that walks the music directory tree and collects track filenames, but the many-to-one relationships in the track-name-to-filename translation mean that I can't fully automate the cddbread reconstruction. Since it could also be useful for people who want to build a "My Music Collection" database, perhaps a good solution would be to add an option that always saves a copy of whichever cddbread.n file was selected, and the Musicbrainz ID as cddbread.discid and mbid.discid in an abcde.metadata directory?